The rapid expansion of the digital economy, exponential data growth, new data protection bills, and the deployment of new applications in India have seen a surge in demand for constructing new data centers across the country. Organizations are deploying computers, storage, applications, and services across the core, edge, and cloud to meet their digital transformation objectives. Outsourcing to third-party data center colocation is a flexible, cost-effective, and sustainable option for businesses. According to IDC, data center colocation services revenue in the Asia/Pacific region stood at around USD 15.4 billion by 2023 and is expected to grow at a CAGR of 7.4% until 2027.
The India market is driven by continued investment in digital infrastructure across hyperscaler data centers, enterprises, and government agencies outsourcing to colocation providers. According to the IDC Data Center Survey 2022, at least 41% of compute and storage assets are deployed in colocation data centers in India. This leads to a massive increase in power consumption by these data centers, resulting in higher demand for cooling systems. Data center cooling systems can consume massive amounts of water, which is a major constraint in India. So, data center users should evaluate suitable cooling solutions that use minimal or no water.
Investments in intelligent power and cooling systems have been the top priority for DC service providers. The growth of Generative AI use cases would drive the demand for AI-designed servers which utilize energy-intensive chips like graphics processing units (GPUs) and tensor processing units (TPUs). This will increase power consumption and result in significant heat densities in data centers.
As the demand for data center infrastructure continues to rise, driven by the growth of generative AI applications, the need for more efficient cooling methods in high-power, highdensity environments becomes increasingly critical. Traditional air-cooling systems often struggle to manage the heat generated by these power-dense racks, leading to the exploration of alternative cooling solutions. In a discussion with analyst firm, IDC, Vertiv revealed the advanced strategies and innovative technologies driving the evolution of data center cooling. Here are the key highlights from their insightful conversation.
IDC’s Question:
What are some of the sustainable data center cooling techniques that are available in the market today? What are some of the differentiated offerings that customers can benefit from?
Vertiv:
There are broadly two methods of data center cooling: air cooling and liquid cooling. Air cooling can be further divided into direct expansion DX-based and chiller water-based systems. DX-based CRAC systems can be used with or without an EconoPhase refrigerant pumping system. The EconoPhase technique uses ambient air to cool refrigerant without water, making it one of the most sustainable data center cooling techniques.
In chilled water-based systems, new adiabatic freecooling chillers with fanwalls are more prominent for large data center applications. With the increase in data center rack load, row-based cooling, rack cooling or closed-room cooling techniques, and liquid cooling are becoming more popular.
There are two types of liquid cooling: single-phase and two-phase. As rack power density in data centers increases, the demand for efficient cooling methods is growing. Liquid cooling is more efficient than air cooling because liquids can absorb and regulate heat better. Additionally, liquid cooling systems take up less space and maximize IT equipment capacity.
IDC’s Question:
How will AI-based advanced monitoring and control software help in tracking the power usage patterns in a data center and applying advanced automation techniques to optimize cooling system operations to consume less energy?
Vertiv:
Most data centers currently use preventive monitoring solutions for power and cooling operations. With the help of artificial intelligence, we can advance to predictive monitoring solutions. These solutions provide periodic checks on the health of key components and generate alarms if any system reports errors, such as filter replacement, fan performance issues, or compressor tracking, etc. This proactive approach helps to identify potential failures early, allowing for preemptive actions to avoid risks that could impact operations. Additionally, monitoring and measuring power usage can reveal consumption patterns and highlight areas where power can be saved.
Advanced monitoring and control applications can provide precise and regular performance data. For example, the Vertiv™ Liebert® iCOM™-S optimizes data center cooling systems through thermal monitoring and control. The Vertiv™ CBM solution uses AI and machine learning to generate health scores and detect potential failures, improving maintenance schedules and reducing downtime. The Vertiv™ Environet™ solution enables comprehensive data collection for efficient data center management, enhancing visibility and control.
IDC’s Question:
As AI adoption grows in India, what are some of the future trends in data center cooling techniques expected in the future?
Vertiv:
As AI and Gen AI offer immense potential, their high-power demand will pose a significant heat dissipation challenge. According to IDC research, total AI IT systems (servers, storage, and networking) would contribute to 12% of the global data center energy consumption by 2024. This is expected to increase to 18% by 2028. Traditional air-cooling methods depend on air to remove heat from the servers. However, AI loads would create a dense section of high heat within the servers, thus increasing heat dissipation at rack level.
Air cooling methods may be insufficient for removing substantial heat due to poor air conductivity. To address this, data centers are adopting liquid cooling, which is more efficient at eliminating excess heat. Liquid cooling is divided into two main approaches: direct-to-chip and immersion cooling. Currently, single-phase direct-to-chip liquid cooling is more common, but as server loads increase, two-phase liquid cooling will become more important. Vertiv offers both cooling methods.
IDC’s Question:
Considering the increasing power density requirements for AI workloads, what are the solutions available to mitigate the power distribution and architecture challenges?
Vertiv:
A hybrid power infrastructure approach is beneficial in colocation scenarios with distributed workloads. By using a modular design for flexible scaling and efficient power distribution, power infrastructure can be segmented to support traditional workloads while dedicated systems, like OCP, handle spiked AI workloads. This prevents oversizing the entire colocation data center and maximizes energy density, scalability, and operational flexibility.
